SKOS Tutorial Catch Mark van Assem, Antoine Isaac Vrije Universiteit Amsterdam Based on slides by Alistair Miles CCLRC Rutherford Appleton Laboratory Semantic Web Best Practices and Deployment

Intro SKOS –“Simple Knowledge Organisation System(s)” –Simple, extensible, machine-understandable representation for “concept schemes” Thesauri Classification Schemes Taxonomies Subject Headings Other types of ‘controlled vocabulary’…

SKOS Development Developed by W3C’s Semantic Web Best Practices-WG Draft for Working Group Note Design: public, consensus-driven, open community, Input from actual vocabulary maintainers

Motivation Semantic Web technology can help improve search facilities and reuse: 1.Concept-based search instead of text- based search 2. Reuse each other’s concept definitions 3. Search across (institution) boundaries 4. Standard software

1. Concept Search Painter Domenikos Theotocopoulos = “El Greco” (nickname) Some indexers use “El Greco”, others “D. Theotocopoulos” Searching for “El Greco” does not give all results Solution: one concept with different lexical labels.

Example N.B.: vocabulary with identifiers for preferred terms and indexing with identifiers accomplishes this

2. Reuse Reuse existing concept “El Greco”  Req. 1: one “exchange syntax”  Req. 2: “point” at other concepts

3. Search Across Boundaries Search for concept “El Greco” returns paintings from both institutions  Same requirements

4. Standard Software If all concept schemes use same “exchange syntax” and “structure”, standardized software can be built to: –Display/browse concept scheme –Annotate with concept scheme –Integrate data from 2 institutions using standard concept schemes (“search across boundaries”)  Req. 3: Similar structures (graphs) in exchange syntax

Why SKOS helps SKOS uses RDF –sharing “graphs” in distributed environment (intranet/internet) –Uses URIs for “pointing” (identifying) –Easy to extend by anyone for specific purposes  “exchange syntax”  “Point at concept” SKOS: set of classes and properties to describe concept schemes -Produce “similar graphs”  “Same structures”/ clear what graph means Disadvantage: unusual concept schemes don’t fit into SKOS (original structure too complex)

Quick RDF: a ‘Statement’ A.K.A. a ‘Triple’

Quick RDF: a ‘Graph’

Quick RDF: exchange syntax RDF Graphs can be exchanged in XML (and other formats) Alternative ways to represent & exchange the same graph Here we only discuss RDF graphs, exchange syntax is “lower-level” technical issue

Controlled Vocabulary Love Strong feelings of attraction towards, and affection for, another adult, or great affection for a friend or family member. Awe A feeling of great respect sometimes mixed with fear or surprise. Joy A feeling of bliss and great happiness.

Converting into SKOS graph 1.Identify 2.Describe 3.Publish

Identify Step 1: Identify concepts…

Describe Step 2: Describe…

Publish Step 3: Publish… –Put the file on a web server for programs to download & process –Put the file on special RDF server on which you can query with SQL-like language: Select * from … Where …

Thesaurus (USE/UF) Love( preferred term) UF Affection Affection ( non-preferred term) USE Love (“USE” directs user from non-pref term to pref-term that should be used in indexing and search)

Lexical Labels

Thesaurus (BT/NT) Love BT Emotion(“ BT” = Broader Term) Emotion NT Love(“ NT” = Narrower Term ) NT Awe NT Joy ( BT/NT only between preferred terms )

Broader/Narrower

Thesaurus (RT) Love RT Beauty (“RT” = Related Term) Beauty RT Love (RT only between preferred terms)

Related

Story So Far… Basic Structure –skos:Concept Lexical Labelling –skos:prefLabel, skos:altLabel Documentation –skos:definition Semantic Relations –skos:broader, skos:narrower, skos:related

More Documentation Properties skos:note e.g. ‘I’m going bananas’ skos:definition e.g. ‘A long curved fruit with a yellow skin and soft, sweet white flesh inside.’ skos:example e.g. ‘A bunch of bananas.’ skos:scopeNote e.g. ‘Only use for the western family of bananas’ skos:historyNote e.g. ‘Introduced 1986.’

Concept Schemes Organise a set of concepts into a concept scheme Add metadata about the scheme –Title –Rights –creator

Concept Scheme

Top Concepts

Subject Indexing One of the main uses of concept scheme is to index documents, pictures, … skos:subject

Spotted Bowerbird

Subject

Node Labels in Hierarchy milk (node label) buffalo milk cow milk goat milk sheep milk (Organize terms into “subcategories” to help users find relevant term; “guide terms”; node label itself not meant for indexing)

Representation in SKOS

Story So Far… Documentation Properties –skos:note, skos:definition, skos:example, skos:scopeNote, skos:historyNote Concept Schemes –skos:ConceptScheme, skos:hasTopConcept, Subject Indexing –skos:subject Node Labels –skos:Collection, skos:member More properties not shown here

Extensions SKOS Core can be extended by refining the classes and properties of the SKOS RDF Schema E.g. North-Holland BT Netherlands is a part-of relationship

Example

Exercise Groups of six, separated into group A and group B Given a “concept scheme” (text), create SKOS graph for it Groups A: give SKOS graph (not the text) to group B (and vice versa) Re-create “concept scheme” (text) from SKOS graph

SKOS Graph Legend Concept in concept scheme String Arrow types: broader, narrower, related, broaderPartOf, … prefLabel, altLabel, scopeNote, definition, historyNote, … Introduce new “arrow types” if required “text”

Possible Exercise Solution (Brinkman)

Possible Exercise Solution (NBC)

Conclusions from exercise Different syntax/structures cause problems SKOS graph makes graphs for concept schemes more uniform Not trivial how to convert original source into SKOS graph But once agreement on that  interoperability

Last Point Do we expect everyone to change to SKOS? No, internal formats and SKOS can co-exist Export to SKOS RDF for interoperability Right tool for the right job!

Links SKOS Core Homepage SKOS Core Guide SKOS Core Vocabulary Specification Mailing list